CN102567304A - Filtering method and device for network malicious information - Google Patents
Filtering method and device for network malicious information Download PDFInfo
- Publication number
- CN102567304A CN102567304A CN2010106211421A CN201010621142A CN102567304A CN 102567304 A CN102567304 A CN 102567304A CN 2010106211421 A CN2010106211421 A CN 2010106211421A CN 201010621142 A CN201010621142 A CN 201010621142A CN 102567304 A CN102567304 A CN 102567304A
- Authority
- CN
- China
- Prior art keywords
- filtered
- information
- model information
- text message
- user feedback
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 60
- 238000001914 filtration Methods 0.000 title claims abstract description 36
- 238000011160 research Methods 0.000 claims abstract description 51
- 239000000463 material Substances 0.000 claims description 66
- 230000008878 coupling Effects 0.000 claims description 31
- 238000010168 coupling process Methods 0.000 claims description 31
- 238000005859 coupling reaction Methods 0.000 claims description 31
- 230000010365 information processing Effects 0.000 abstract description 3
- 230000008569 process Effects 0.000 description 18
- 238000012549 training Methods 0.000 description 7
- 238000005516 engineering process Methods 0.000 description 5
- 238000006243 chemical reaction Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 230000003203 everyday effect Effects 0.000 description 3
- 238000002360 preparation method Methods 0.000 description 3
- 238000013145 classification model Methods 0.000 description 2
- 239000003814 drug Substances 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 238000005194 fractionation Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000003058 natural language processing Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- PCTMTFRHKVHKIS-BMFZQQSSSA-N (1s,3r,4e,6e,8e,10e,12e,14e,16e,18s,19r,20r,21s,25r,27r,30r,31r,33s,35r,37s,38r)-3-[(2r,3s,4s,5s,6r)-4-amino-3,5-dihydroxy-6-methyloxan-2-yl]oxy-19,25,27,30,31,33,35,37-octahydroxy-18,20,21-trimethyl-23-oxo-22,39-dioxabicyclo[33.3.1]nonatriaconta-4,6,8,10 Chemical compound C1C=C2C[C@@H](OS(O)(=O)=O)CC[C@]2(C)[C@@H]2[C@@H]1[C@@H]1CC[C@H]([C@H](C)CCCC(C)C)[C@@]1(C)CC2.O[C@H]1[C@@H](N)[C@H](O)[C@@H](C)O[C@H]1O[C@H]1/C=C/C=C/C=C/C=C/C=C/C=C/C=C/[C@H](C)[C@@H](O)[C@@H](C)[C@H](C)OC(=O)C[C@H](O)C[C@H](O)CC[C@@H](O)[C@H](O)C[C@H](O)C[C@](O)(C[C@H](O)[C@H]2C(O)=O)O[C@H]2C1 PCTMTFRHKVHKIS-BMFZQQSSSA-N 0.000 description 1
- 244000188472 Ilex paraguariensis Species 0.000 description 1
- 206010028916 Neologism Diseases 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000009792 diffusion process Methods 0.000 description 1
- 230000009977 dual effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000008713 feedback mechanism Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000008521 reorganization Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/10—Text processing
- G06F40/166—Editing, e.g. inserting or deleting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/335—Filtering based on additional data, e.g. user or group profiles
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/205—Parsing
Abstract
The invention discloses a filtering method and device for network malicious information and relates to the technical field of computer information processing and information filtering. The filtering method disclosed by the embodiment of the invention comprises the following steps of: obtaining text information to be filtered, system pre-research model information and user feedback model information; pre-treating the text information to be filtered; carrying out characteristic information matching on the pre-treated text information to be filtered and the system pre-research model information to obtain a first matched result; carrying out the characteristic information matching on the pre-treated text information to be filtered and user feedback model information to obtain a second matched result; and carrying out filtering treatment on the text information to be filtered according to the first matched result and the second matched result. With the adoption of the filtering method and device of the network malicious information, disclosed by the embodiment of the invention, the automatic filtering performance for the malicious information is improved and system information can be automatically updated.
Description
Technical field
The present invention relates to computer information processing and information filtering technical field, relate in particular to a kind of based on the filter method and the device of statistics with the network flame of rule.
Background technology
Along with Internet fast development, the information velocity of propagation is also accelerated thereupon.Because the content on the internet is very different; For example: advertisement, pornographic, violence and reaction are that main flame all is difficult to stop; And spread with more hidden mode gradually, therefore, suppress the diffusion of flame and purify the space, internet just to seem very important.For the data message of magnanimity in the internet,, then need expend huge manpower and materials if adopt artificial method to remove to filter the flame on the internet.Therefore, become the focus of Recent study based on the automatic fitration technology of the flame of internet content.
At present, the flame automatic fitration technology based on internet content adopts following dual mode usually:
(1) based on the filter method of keyword matching; This method is taked the accurately strategy of coupling in decision process, filter out the text that key word occurs.Adopt the flame speed of this method filtration internet content fast, simple to operation.
(2) based on the filter method of textual classification model of statistics; Bad text filtering model based on statistics in this method is one two types text classification problem in essence, and text classification is the research emphasis direction of natural language processing field, has a large amount of classical models can be for reference.Textual classification model based on statistics should be the pretty good method of effect from point of theory, but performance is undesirable in practical application, and the erroneous judgement situation is very outstanding, and the main cause analysis is following:
(1) forward and negative sense language material are unbalanced.Wherein, the forward language material has only comprised a small amount of classification, and for example: advertisement, pornographic, violence, reaction and the flame that the user was concerned about are main.The negative sense language material has then comprised a large amount of classifications, for example: can be divided into according to content of text: economy, physical culture, politics, medicine, art, history, politics, culture, environment, traffic, computing machine, education, military affairs or the like.
(2) performance of the content of flame has very big polytrope and disguise.The publisher often has a mind to avoid everyday words, replaces, as: phonetically similar word splits word, non-Chinese character noise, breviary phenomenon, neologisms etc.
(3) user-oriented dictionary only provides keyword accurate matching way, causes the machinery of decision method and dumb.And the semantic tendency property of single keyword is not representative, and False Rate is high.Such as, when appearing in the context environmental simultaneously, " freely " and " invoice " have more persuasion property than single " invoice ".
(4) some traditional Chinese information processing ways and not being suitable for based on the flame of text classification are filtered.As use the stop word of certain scale; Include only vocabulary more than the double word etc. like characteristic item.
(5) lack unified model, carry out synthetic filter comprising flames such as advertisement, pornographic, violence, reaction.
State in realization in the process based on the flame automatic fitration of internet content technology, the inventor finds in the prior art that flame automatic fitration performance can't satisfy the filtration needs of current internet, and can't realize automatic renewal.
Summary of the invention
The embodiment of the invention provides a kind of filter method and device of network flame, and for achieving the above object, embodiments of the invention adopt following technical scheme:
A kind of filter method of network flame comprises:
Obtain text message to be filtered, system's beforehand research model information and user feedback model information;
Said text message to be filtered is carried out pre-service;
Said pretreated text message to be filtered and said system beforehand research model information are carried out the characteristic information coupling, provide first matching result;
Said pretreated text message to be filtered and said user feedback model information are carried out the characteristic information coupling, provide second matching result;
According to said first matching result and said second matching result, said text message to be filtered is carried out filtration treatment.
A kind of filtration unit of network flame comprises:
Information acquisition unit is used to obtain text message to be filtered, system's beforehand research model information and user feedback model information;
Pretreatment unit is used for said text message to be filtered is carried out pre-service;
First matching unit is used for said pretreated text message to be filtered and said system beforehand research model information are carried out the characteristic information coupling, provides first matching result;
Second matching unit is used for said pretreated text message to be filtered and said user feedback model information are carried out the characteristic information coupling, provides second matching result;
Filter element is used for according to said first matching result and said second matching result said text message to be filtered being carried out filtration treatment.
The filter method and the device of the network flame that the embodiment of the invention provides are through obtaining text message to be filtered, system's beforehand research model information and user feedback model information; Said text message to be filtered is carried out pre-service; Said pretreated text message to be filtered and said system beforehand research model information are carried out the characteristic information coupling, provide first matching result; Said pretreated text message to be filtered and said user feedback model information are carried out the characteristic information coupling, provide second matching result; According to said first matching result and said second matching result, said text message to be filtered is carried out filtration treatment.Owing to adopted twice coupling to carry out system filtration among the present invention,, thereby improved the performance of system so the accuracy of system's automatic fitration flame is higher; Also, make field feedback to be applied to timely in the process of system's automatic fitration flame, thereby realized the function that system model information is upgraded automatically because the embodiment of the invention has adopted the user feedback model information to carry out the filtration of flame.
Description of drawings
The filter method process flow diagram of a kind of network flame that Fig. 1 provides for the embodiment of the invention;
The filter method process flow diagram of the another kind of network flame that Fig. 2 provides for the embodiment of the invention;
The filter apparatus configuration synoptic diagram of a kind of network flame that Fig. 3 provides for the embodiment of the invention;
The filter apparatus configuration synoptic diagram of the another kind of network flame that Fig. 4 provides for the embodiment of the invention.
Embodiment
The filter method and the device of a kind of network flame that the embodiment of the invention is provided below in conjunction with accompanying drawing are described in detail.
Of Fig. 1, the filter method of a kind of network flame that provides for the embodiment of the invention; This method comprises:
101: obtain text message to be filtered, system's beforehand research model information and user feedback model information;
102: said text message to be filtered is carried out pre-service;
103: said pretreated text message to be filtered and said system beforehand research model information are carried out the characteristic information coupling, provide first matching result;
104: said pretreated text message to be filtered and said user feedback model information are carried out the characteristic information coupling, provide second matching result;
105:, said text message to be filtered is carried out filtration treatment according to said first matching result and said second matching result.
Of Fig. 2, the filter method of the another kind of network flame that provides for the embodiment of the invention, this method comprises:
201: obtain the language material of said system beforehand research model information and the language material of said user feedback model information.Wherein, the language material of said user feedback model information can comprise: user feedback language material and/or be filtered language material.The selection of the corpus of common said system beforehand research model and said user feedback model is divided into forward language material and negative sense language material; For example: the collection of the flame content text of the preparation of forward language material can mainly comprise: content text such as advertisement, pornographic, violence, reaction, totally 10000 pieces; The collection of the non-flame content text of the preparation of negative sense language material mainly comprises the main text categories of task; Like economy, politics, physical culture, culture, medicine, traffic, environment, military affairs, literature and art, history, computing machine, education, law, house property, science and technology, automobile, the talent, amusement etc., totally 30000 pieces.
Need to prove that in the collection process of said corpus, it is unbalanced positive and negative language material often to occur; The language material wide range of a classification, another classification language material scope is then less relatively.Solution among the present invention is to allow this unbalanced language material to distribute, and is to demand perfection not ask amount for the preparation strategy of the very big classification of language material scope.
202: obtain text message to be filtered, system's beforehand research model information and user feedback model information;
203: said text message to be filtered is carried out pre-service;
This step specifically comprises: said text message to be filtered is carried out cutting handle; For example: according to punctuate and common speech language material is made pauses in reading unpunctuated ancient writings, common speech is meant commonly used and to judging insignificant vocabulary, as " ", " " etc., but " you " are common in the forward language material, and " I " am common in the negative sense language material, am not suitable for as everyday words.
It should be noted that include list commonly used in the natural language processing is not suitable for as vocabulary commonly used.Usually can adopt upright intelligence to think 4.0 pairs of language materials of participle and carry out participle and part-of-speech tagging work.Cutting unit after said cutting is handled is the minimum processing unit of follow-up work.
Add up the candidate feature item quantity after said cutting is handled.For example: to the wherein non-Chinese character part quantity of cutting unit statistics after the said cutting processing; As: said cutting unit adds up to N1, and non-Chinese character part is N2, if N2/N1 greater than threshold value, judges that then the pairing text message to be filtered of this candidate feature item is a flame.According to being to contain a large amount of noise characters in this information, possibly be rubbish texts such as advertisement; Perhaps; Add up contact methods such as network address in the said cutting unit, phone, mailbox, QQ and quantity num (ad) occurs; This type of information is usually used in the advertisement, and gives default-weight
204: said pretreated text message to be filtered and said system beforehand research model information are carried out the characteristic information coupling, provide first matching result; This step specifically can comprise:
2041: obtain said pretreated text message to be filtered and said system beforehand research model information; Said system beforehand research model information comprises: rule index storehouse and said system beforehand research aspect of model item information; Wherein, the generative process in the user policy index database in the said rule index storehouse and user's keyword index storehouse is following:
Step S1: keyword resolution; Said keyword resolution method is: at first, index built in the Chinese phonetic alphabet of everyday character, generate the index of whole keyword according to the Chinese phonetic alphabet index of each word in the keyword; Then, again each word in the keyword is carried out structural fractionation, according to split result recurrence reorganization keyword; At last, keyword index and fractionation set are formed key-value pair, preserve all analysis results and generate user's keyword index storehouses.Behind " Falun Gong " keyword resolution, can generate an index value, and multiple split result is arranged, specifically can comprise " three go car logical sequence skill ", " Fa Chelun merit " or the like.
Step S2: syntax parsing; Computing machine resolves to rule syntax can forms of treatment.Said rule syntax comprises: AND, OR, NEAR, NOT.Like " A ANDB ", wherein A and B are keywords to be resolved, and the AND syntactic representation is in context environmental, and when A and B occurred simultaneously, this rule was matched to merit.Keyword and rule syntax are formed key-value pair, preserve all analysis results and generate the user policy index database.
It should be noted that the above index database rule can be user configured rule, all right system intialization rule; The above step is user's configuration rule to be resolved generate corresponding index database process, and this index database can be optimized following matching process.
2042: said pretreated text message to be filtered and said system beforehand research model information are mated, obtain characteristic item; Wherein, said system beforehand research model information comprises: rule index storehouse and said system beforehand research aspect of model item information; The process that this step is obtained system's beforehand research aspect of model item information specifically can for:
Step S1 forms the speech string as the candidate feature item with said cutting unit; For example:
(1) continuous cutting unit combination is become the speech string.For the cutting unit in every, since the 1st cutting unit, combination window is N to the maximum, makes up.Like orderly cutting unit " ABCD ", maximized window is 3, and the combination that then generates the speech string has 9 kinds: ABC, BCD, AB, BC, CD, A, B, C, D.
(2) discrete cutting unit combination is become the speech string.Speech string to the generation in (1) calculates Chinese phonetic alphabet index, matees in user's keyword index storehouse that the step S1 in the foundation said 2041 generates.If the successful set of coupling is arranged, statistical match success quantity num (user); Then, mate in the user policy index database that the step S2 in the foundation said 2041 generates again,, generate a speech string for discrete cutting unit if mate successfully.Like 9 speech strings in (1), if in user's keyword index storehouse, mate successfully two speech string A, D.Regular in the user policy index database " A NEAR2 D " then generates new characteristic item AD.Here 2 represent the distance of A and D to be no more than 2.The statistical match that adds up success quantity num (user) gives default-weight
Step S2 carries out the frequency to said candidate feature item and filters; Concretely; Be exactly the occurrence number of statistics candidate feature item in corpus, filter as index, the candidate feature item of the frequency more than or equal to threshold value kept with the frequency; Candidate feature item less than threshold value is rejected, and can adjust threshold value the scope that keeps is controlled.
Step S3 carries out the frequency to said candidate feature item and refilters; Concrete filter process comprises:
At first, irrational frequency is reappraised, such as, all be the situation of AB when B occurring as if all, the then frequency vanishing of B.The frequency reappraises formula:
Wherein, a representation feature item; The word frequency of f (a) expression a; B has represented to comprise the long string characteristic item of a; The set of
expression b;
representes set sizes.
Then, filter once more as index, the candidate feature item of the frequency more than or equal to threshold value kept, reject, can adjust threshold value, the scope that keeps is controlled less than the candidate feature item of threshold value with the frequency after reappraising.
Step S4: said candidate feature item is selected automatically, thereby extracted characteristic item.Concretely; The candidate feature item that to be exactly this step get access to the forward language material from said step S3 candidate feature item and negative sense language material obtain from said step S3 merges; Therefore merging these candidate feature items of back has two word frequency, respectively the corresponding forward frequency and the negative sense frequency.Adopt statistical chi amount to carry out the automatic selection of characteristic item, keep the maximum top n candidate feature item of chi-square value as final characteristic item information.Chi amount formula is:
Wherein the implication of A, B, C, D, N is following:
K only gets 0 or 1 in the table, represents two kinds, i.e. forward classification and negative sense classification.
Need to prove that said characteristic item comprises monosyllabic word and multi-character words.Monosyllabic word is bigger to the judgement influence of negative sense text.The content of forum's text message particularly, the cutting unit of individual character is more common, if do not consider individual character, the negative sense text is easy to cause erroneous judgement.
2043: the language material information score of adding up said characteristic item; In step S4, preserved the frequency of said characteristic item; And each characteristic item all has two frequencys, represents the forward frequency and the negative sense frequency respectively, such as; The forward frequency of " invoice " will be far longer than the negative sense frequency, because " invoice " more is common in the flame of advertisement.Regard the forward frequency of each characteristic item the forward weight of characteristic item as, the negative sense frequency of each characteristic item is regarded as the negative sense weight of characteristic item.For all characteristic items, align the negative sense weight respectively and carry out normalization, like this, weighted value just has comparative sense.Normalized formula is:
Two types of language material training obtain because characteristic item that generates and weight thereof are according to the pre-prepd standard of system, preserve and generate the result as system's beforehand research aspect of model item information.
Said pretreated text message to be filtered and said system beforehand research aspect of model item information are carried out the characteristic information coupling, obtain text feature item information to be filtered, calculate said characteristic item information forward score, its computing formula is:
Calculate said characteristic item information negative sense score, its computing formula is:
Simultaneously, consider num (ad) and num (user), aforementioned calculation formula right side is changed to:
2044:, judge whether the pairing text message to be filtered of said characteristic item is flame according to said language material information score; If
then system's beforehand research model information judges that this pending text is bad text; If
then this model lost efficacy, judge failure: then system's beforehand research model information judges that this pending text is a normal text as if
.
2045:, provide said first matching result according to judged result.
205: said pretreated text message to be filtered and said user feedback model information are carried out the characteristic information coupling, provide second matching result; The said flow process of flow process that this step specifically can comprise and step 204 is roughly the same.
Need to prove that the said process of obtaining the user feedback model information is the selection of corpus in the step 201 with the place that obtains the process main difference of system's beforehand research model information.The source of the corpus of said user feedback model information can also comprise following two aspects:
(1) user feedback mechanisms.The user finds to judge the information that goes wrong in the real experiences process, mainly is the situation that flame is judged to be normal information, and system is reported an error, and system receives the user answer as the feedback language material.
(2) decision model mechanism.Pending text gets into the flame determination flow of step 206, and output is to the result of determination of the text.Two kinds of situation that the result comprises, promptly bad text or normal text.According to judging that the confidence level situation determines whether pending text participates in feedback training.
206:, said text message to be filtered is carried out filtration treatment according to said first matching result and said second matching result.Concretely, judge exactly whether said first matching result is consistent with the result of determination of said second matching result, i.e. the result of determination of system's beforehand research model information and user feedback model information.If judgement is identical, be all flame text or normal information text, then the result of determination confidence level is bigger, can be used for feedback training; If judge difference, then the result of determination confidence level has loss, but if take comparatively strict filtering policy, then filters this text, but be not useable for feedback training; If wherein there is a model to lose efficacy, then the result is according to the result of determination of residue model, and thinks that certain confidence level is arranged, and can be used for feedback training; If two models all lost efficacy, then return the sign that lost efficacy, be not useable for feedback training.
After it should be noted that the decision process of a text message to be filtered of every completion, this method can also comprise:
The language material quantity of obtaining said user feedback model information with and corresponding threshold; Concretely, statistics can be used for the language material quantity of feedback training exactly, judges whether said language material quantity exceeds its corresponding threshold value.
According to the language material quantity of said user feedback model information with and corresponding threshold, said user feedback model information is upgraded.If language material quantity greater than threshold value, is then trained the feedback language material again, upgrade the user feedback model information.The size of adjustment threshold value can be adjusted the update cycle.
As shown in Figure 3, the filtration unit of a kind of network flame that provides for the embodiment of the invention; This device comprises:
First matching unit 303 is used for said pretreated text message to be filtered and said system beforehand research model information are carried out the characteristic information coupling, provides first matching result;
As shown in Figure 4, the filtration unit of a kind of network flame that provides for the embodiment of the invention; This device comprises:
Cutting subelement 4021 is used for that said text message to be filtered is carried out cutting and handles;
First matching unit 403 is used for said pretreated text message to be filtered and said system beforehand research model information are carried out the characteristic information coupling, provides first matching result; This unit specifically can comprise:
Information is obtained subelement 4031, is used to obtain said pretreated text message to be filtered and said system beforehand research model information; Wherein, said system beforehand research model information comprises: rule index storehouse and said system beforehand research aspect of model item information;
The result exports subelement 4035, is used for according to judged result, provides said first matching result.
Information is obtained subelement 4041, is used to obtain said pretreated text message to be filtered and said user feedback model information; Wherein, said user feedback model information comprises: rule index storehouse and said user feedback aspect of model item information;
The result exports subelement 4045, is used for according to judged result, provides said second matching result.
It should be noted that this device also comprises:
Threshold value acquiring unit 406, the language material quantity that is used to obtain said user feedback model information with and corresponding threshold;
Updating block 407, be used for according to the language material quantity of said user feedback model information with and corresponding threshold, said user feedback model information is upgraded.When the language material quantity of the user feedback model information that gets access to when said threshold value acquiring unit reaches its corresponding threshold; Said updating block according to the language material quantity of said user feedback model information with and corresponding threshold, said user feedback model information is upgraded.
The filter method and the device of the network flame that the embodiment of the invention provides are through obtaining text message to be filtered, system's beforehand research model information and user feedback model information; Said text message to be filtered is carried out pre-service; Said pretreated text message to be filtered and said system beforehand research model information are carried out the characteristic information coupling, provide first matching result; Said pretreated text message to be filtered and said user feedback model information are carried out the characteristic information coupling, provide second matching result; According to said first matching result and said second matching result, said text message to be filtered is carried out filtration treatment.Owing to adopted twice coupling to carry out system filtration among the present invention,, thereby improved the performance of system so the accuracy of system's automatic fitration flame is higher; Also because the embodiment of the invention has adopted the user feedback model information to carry out the filtration of flame; Make field feedback to be applied to timely in the process of system's automatic fitration flame, thereby realized the automatic function of upgrading of match information of system.
Description through above embodiment; One of ordinary skill in the art will appreciate that: realize that all or part of step in the foregoing description method is to instruct relevant hardware to accomplish through program; Described program can be stored in the computer read/write memory medium, and this program comprises the step like above-mentioned method embodiment when carrying out; Described storage medium, as: ROM/RAM, magnetic disc, CD etc.
The above; Be merely embodiment of the present invention, but protection scope of the present invention is not limited thereto, any technician who is familiar with the present technique field is in the technical scope that the present invention discloses; Can expect easily changing or replacement, all should be encompassed within protection scope of the present invention.Therefore, protection scope of the present invention should be as the criterion with the protection domain of claim.
Claims (16)
1. the filter method of a network flame is characterized in that, comprising:
Obtain text message to be filtered, system's beforehand research model information and user feedback model information;
Said text message to be filtered is carried out pre-service;
Said pretreated text message to be filtered and said system beforehand research model information are carried out the characteristic information coupling, provide first matching result;
Said pretreated text message to be filtered and said user feedback model information are carried out the characteristic information coupling, provide second matching result;
According to said first matching result and said second matching result, said text message to be filtered is carried out filtration treatment.
2. the filter method of network flame according to claim 1 is characterized in that, this method also comprises:
Obtain the language material of said system beforehand research model information and the language material of said user feedback model information.
3. the filter method of network flame according to claim 2 is characterized in that, the language material of said user feedback model information comprises: user feedback language material and/or be filtered language material.
4. the filter method of network flame according to claim 3 is characterized in that, this method also comprises:
The language material quantity of obtaining said user feedback model information with and corresponding threshold;
According to the language material quantity of said user feedback model information with and corresponding threshold, said user feedback model information is upgraded.
5. according to the filter method of claim 2 or 3 or 4 described network flames, it is characterized in that, said said text message to be filtered carried out pretreated step, comprising:
Said text message to be filtered is carried out cutting to be handled;
Add up the candidate feature item quantity after said cutting is handled.
6. the filter method of network flame according to claim 5 is characterized in that, said said pretreated text message to be filtered and said system beforehand research model information is carried out the characteristic information coupling, provides the first matching result step, comprising:
Obtain said pretreated text message to be filtered and said system beforehand research model information;
Said pretreated text message to be filtered and said system beforehand research model information are mated, obtain characteristic item;
Add up the language material information score of said characteristic item;
According to said language material information score, judge whether the pairing text message to be filtered of said characteristic item is flame;
According to judged result, provide said first matching result.
7. the filter method of network flame according to claim 6 is characterized in that, said said pretreated text message to be filtered and said user feedback model information is carried out the characteristic information coupling, provides the second matching result step, comprising:
Obtain said pretreated text message to be filtered and said user feedback model information;
Said pretreated text message to be filtered and said user feedback model information are mated, obtain characteristic item;
Add up the language material information score of said characteristic item;
According to said language material information score, judge whether the pairing text message to be filtered of said characteristic item is flame;
According to judged result, provide said second matching result.
8. according to the filter method of claim 6 or 7 described network flames, it is characterized in that said system beforehand research model information comprises: rule index storehouse and system's beforehand research aspect of model item information; Said user feedback model information comprises: rule index storehouse and user feedback aspect of model item information.
9. the filter method of network flame according to claim 8 is characterized in that, the rule index storehouse of said system beforehand research model information comprises: the system intialization rule; The rule index storehouse of said user feedback model information comprises: user's configuration rule.
10. the filtration unit of a network flame is characterized in that, comprising:
Information acquisition unit is used to obtain text message to be filtered, system's beforehand research model information and user feedback model information;
Pretreatment unit is used for said text message to be filtered is carried out pre-service;
First matching unit is used for said pretreated text message to be filtered and said system beforehand research model information are carried out the characteristic information coupling, provides first matching result;
Second matching unit is used for said pretreated text message to be filtered and said user feedback model information are carried out the characteristic information coupling, provides second matching result;
Filter element is used for according to said first matching result and said second matching result said text message to be filtered being carried out filtration treatment.
11. the filtration unit of network flame according to claim 10 is characterized in that, said information acquisition unit also is used to obtain the language material of said user feedback model information.
12. the filtration unit of network flame according to claim 11 is characterized in that, the language material of said user feedback model information comprises: user feedback language material and/or be filtered language material.
13. the filtration unit of network flame according to claim 12 is characterized in that, this side's device also comprises:
The threshold value acquiring unit, the language material quantity that is used to obtain said user feedback model information with and corresponding threshold;
Updating block, be used for according to the language material quantity of said user feedback model information with and corresponding threshold, said user feedback model information is upgraded.
14. the filtration unit according to claim 11 or 12 or 13 described network flames is characterized in that, said pretreatment unit comprises:
The cutting subelement is used for that said text message to be filtered is carried out cutting and handles;
The statistics subelement is used to add up the candidate feature item quantity after said cutting is handled.
15. the filtration unit of network flame according to claim 14 is characterized in that, said first matching unit comprises:
Information is obtained subelement, is used to obtain said pretreated text message to be filtered and said system beforehand research model information;
The coupling subelement is used for said pretreated text message to be filtered and said system beforehand research model information are mated, and obtains characteristic item;
The statistics subelement, the language material information score that is used to add up said characteristic item;
Judgment sub-unit is used for according to said language material information score, judges whether the pairing text message to be filtered of said characteristic item is flame;
The result exports subelement, is used for according to judged result, provides said first matching result.
16. the filtration unit of network flame according to claim 15 is characterized in that, said second matching unit comprises:
Information is obtained subelement, is used to obtain said pretreated text message to be filtered and said user feedback model information;
The coupling subelement is used for said pretreated text message to be filtered and said user feedback model information are mated, and obtains characteristic item;
The statistics subelement, the language material information score that is used to add up said characteristic item;
Judgment sub-unit is used for according to said language material information score, judges whether the pairing text message to be filtered of said characteristic item is flame;
The result exports subelement, is used for according to judged result, provides said second matching result.
Priority Applications (5)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201010621142.1A CN102567304B (en) | 2010-12-24 | 2010-12-24 | Filtering method and device for network malicious information |
| PCT/CN2011/084699 WO2012083892A1 (en) | 2010-12-24 | 2011-12-26 | Method and device for filtering harmful information |
| JP2013545039A JP5744228B2 (en) | 2010-12-24 | 2011-12-26 | Method and apparatus for blocking harmful information on the Internet |
| EP11850052.9A EP2657852A4 (en) | 2010-12-24 | 2011-12-26 | Method and device for filtering harmful information |
| US13/997,666 US20140013221A1 (en) | 2010-12-24 | 2011-12-26 | Method and device for filtering harmful information |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201010621142.1A CN102567304B (en) | 2010-12-24 | 2010-12-24 | Filtering method and device for network malicious information |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102567304A true CN102567304A (en) | 2012-07-11 |
| CN102567304B CN102567304B (en) | 2014-02-26 |
Family
ID=46313198
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201010621142.1A Expired - Fee Related CN102567304B (en) | 2010-12-24 | 2010-12-24 | Filtering method and device for network malicious information |
Country Status (5)
| Country | Link |
|---|---|
| US (1) | US20140013221A1 (en) |
| EP (1) | EP2657852A4 (en) |
| JP (1) | JP5744228B2 (en) |
| CN (1) | CN102567304B (en) |
| WO (1) | WO2012083892A1 (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103246641A (en) * | 2013-05-16 | 2013-08-14 | 李营 | Text semantic information analyzing system and method |
| CN103729384A (en) * | 2012-10-16 | 2014-04-16 | 中国移动通信集团公司 | Information filtering method, system and device |
| WO2015062377A1 (en) * | 2013-11-04 | 2015-05-07 | 北京奇虎科技有限公司 | Device and method for detecting similar text, and application |
| CN105653649A (en) * | 2015-12-28 | 2016-06-08 | 福建亿榕信息技术有限公司 | Identification method and device of low-proportion information in mass texts |
| CN107239447A (en) * | 2017-06-05 | 2017-10-10 | 厦门美柚信息科技有限公司 | Junk information recognition methods and device, system |
Families Citing this family (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103514227B (en) * | 2012-06-29 | 2016-12-21 | 阿里巴巴集团控股有限公司 | A kind of method and device of more new database |
| KR20140025113A (en) * | 2012-08-21 | 2014-03-04 | 한국전자통신연구원 | High speed decision apparatus and method for objectionable contents |
| US9773182B1 (en) * | 2012-09-13 | 2017-09-26 | Amazon Technologies, Inc. | Document data classification using a noise-to-content ratio |
| CN103886026B (en) * | 2014-02-25 | 2017-09-05 | 厦门客来点信息科技有限公司 | Clothing matching method based on personal feature |
| CN104281696B (en) * | 2014-10-16 | 2017-09-15 | 江西师范大学 | A kind of personalized distribution method of the spatial information of active |
| CN105183894B (en) * | 2015-09-29 | 2020-03-10 | 百度在线网络技术（北京）有限公司 | Method and device for filtering website internal links |
| CN105528404A (en) * | 2015-12-03 | 2016-04-27 | 北京锐安科技有限公司 | Establishment method and apparatus of seed keyword dictionary, and extraction method and apparatus of keywords |
| CN106874253A (en) * | 2015-12-11 | 2017-06-20 | 腾讯科技（深圳）有限公司 | Recognize the method and device of sensitive information |
| US10498752B2 (en) | 2016-03-28 | 2019-12-03 | Cisco Technology, Inc. | Adaptive capture of packet traces based on user feedback learning |
| CN106339429A (en) * | 2016-08-17 | 2017-01-18 | 浪潮电子信息产业股份有限公司 | Method, device and system for realizing intelligent customer service |
| CN108038245A (en) * | 2017-12-28 | 2018-05-15 | 中译语通科技（青岛）有限公司 | It is a kind of based on multilingual data digging method |
| CN109597987A (en) * | 2018-10-25 | 2019-04-09 | 阿里巴巴集团控股有限公司 | A kind of text restoring method, device and electronic equipment |
| CN110633466B (en) * | 2019-08-26 | 2021-01-19 | 深圳安巽科技有限公司 | Short message crime identification method and system based on semantic analysis and readable storage medium |
| CN112749565A (en) * | 2019-10-31 | 2021-05-04 | 华为终端有限公司 | Semantic recognition method and device based on artificial intelligence and semantic recognition equipment |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5987457A (en) * | 1997-11-25 | 1999-11-16 | Acceleration Software International Corporation | Query refinement method for searching documents |
| US20040167964A1 (en) * | 2003-02-25 | 2004-08-26 | Rounthwaite Robert L. | Adaptive junk message filtering system |
| CN101477544A (en) * | 2009-01-12 | 2009-07-08 | 腾讯科技（深圳）有限公司 | Rubbish text recognition method and system |
| CN101639824A (en) * | 2009-08-27 | 2010-02-03 | 北京理工大学 | Text filtering method based on emotional orientation analysis against malicious information |
| CN101794303A (en) * | 2010-02-11 | 2010-08-04 | 重庆邮电大学 | Method and device for classifying text and structuring text classifier by adopting characteristic expansion |
| CN101877704A (en) * | 2010-06-02 | 2010-11-03 | 中兴通讯股份有限公司 | Network access control method and service gateway |
| CN101894102A (en) * | 2010-07-16 | 2010-11-24 | 浙江工商大学 | Method and device for analyzing emotion tendentiousness of subjective text |
| CN101908055A (en) * | 2010-03-05 | 2010-12-08 | 黑龙江工程学院 | Method for setting information classification threshold for optimizing lam percentage and information filtering system using same |
Family Cites Families (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5867799A (en) * | 1996-04-04 | 1999-02-02 | Lang; Andrew K. | Information system and method for filtering a massive flow of information entities to meet user information classification needs |
| AU2000233633A1 (en) * | 2000-02-15 | 2001-08-27 | Thinalike, Inc. | Neural network system and method for controlling information output based on user feedback |
| US7543053B2 (en) * | 2003-03-03 | 2009-06-02 | Microsoft Corporation | Intelligent quarantining for spam prevention |
| US7813482B2 (en) * | 2005-12-12 | 2010-10-12 | International Business Machines Corporation | Internet telephone voice mail management |
| US8463810B1 (en) * | 2006-06-01 | 2013-06-11 | Monster Worldwide, Inc. | Scoring concepts for contextual personalized information retrieval |
| US8020206B2 (en) * | 2006-07-10 | 2011-09-13 | Websense, Inc. | System and method of analyzing web content |
| WO2008021244A2 (en) * | 2006-08-10 | 2008-02-21 | Trustees Of Tufts College | Systems and methods for identifying unwanted or harmful electronic text |
| CN101166159B (en) * | 2006-10-18 | 2010-07-28 | 阿里巴巴集团控股有限公司 | A method and system for identifying rubbish information |
| KR100815530B1 (en) * | 2007-07-20 | 2008-04-15 | (주)올라웍스 | Method and system for filtering obscene contents |
| US8965888B2 (en) * | 2007-10-08 | 2015-02-24 | Sony Computer Entertainment America Llc | Evaluating appropriateness of content |
| JP5032286B2 (en) * | 2007-12-10 | 2012-09-26 | 株式会社ジャストシステム | Filtering processing method, filtering processing program, and filtering apparatus |
| EP2071339A3 (en) * | 2007-12-12 | 2015-05-20 | Sysmex Corporation | System for providing animal test information and method of providing animal test information |
| US8850571B2 (en) * | 2008-11-03 | 2014-09-30 | Fireeye, Inc. | Systems and methods for detecting malicious network content |
| US20140108156A1 (en) * | 2009-04-02 | 2014-04-17 | Talk3, Inc. | Methods and systems for extracting and managing latent social networks for use in commercial activities |
| US8849725B2 (en) * | 2009-08-10 | 2014-09-30 | Yahoo! Inc. | Automatic classification of segmented portions of web pages |
| CN101702167A (en) * | 2009-11-03 | 2010-05-05 | 上海第二工业大学 | Method for extracting attribution and comment word with template based on internet |
-
2010
- 2010-12-24 CN CN201010621142.1A patent/CN102567304B/en not_active Expired - Fee Related
-
2011
- 2011-12-26 WO PCT/CN2011/084699 patent/WO2012083892A1/en active Application Filing
- 2011-12-26 EP EP11850052.9A patent/EP2657852A4/en not_active Ceased
- 2011-12-26 US US13/997,666 patent/US20140013221A1/en not_active Abandoned
- 2011-12-26 JP JP2013545039A patent/JP5744228B2/en not_active Expired - Fee Related
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5987457A (en) * | 1997-11-25 | 1999-11-16 | Acceleration Software International Corporation | Query refinement method for searching documents |
| US20040167964A1 (en) * | 2003-02-25 | 2004-08-26 | Rounthwaite Robert L. | Adaptive junk message filtering system |
| CN101477544A (en) * | 2009-01-12 | 2009-07-08 | 腾讯科技（深圳）有限公司 | Rubbish text recognition method and system |
| CN101639824A (en) * | 2009-08-27 | 2010-02-03 | 北京理工大学 | Text filtering method based on emotional orientation analysis against malicious information |
| CN101794303A (en) * | 2010-02-11 | 2010-08-04 | 重庆邮电大学 | Method and device for classifying text and structuring text classifier by adopting characteristic expansion |
| CN101908055A (en) * | 2010-03-05 | 2010-12-08 | 黑龙江工程学院 | Method for setting information classification threshold for optimizing lam percentage and information filtering system using same |
| CN101877704A (en) * | 2010-06-02 | 2010-11-03 | 中兴通讯股份有限公司 | Network access control method and service gateway |
| CN101894102A (en) * | 2010-07-16 | 2010-11-24 | 浙江工商大学 | Method and device for analyzing emotion tendentiousness of subjective text |
Non-Patent Citations (1)
| Title |
|---|
| 马建国 等: "信息过滤技术及Visual J++实现", 《系统工程与电子技术》, vol. 26, no. 3, 31 March 2004 (2004-03-31), pages 382 - 385 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103729384A (en) * | 2012-10-16 | 2014-04-16 | 中国移动通信集团公司 | Information filtering method, system and device |
| CN103729384B (en) * | 2012-10-16 | 2017-02-22 | 中国移动通信集团公司 | information filtering method, system and device |
| CN103246641A (en) * | 2013-05-16 | 2013-08-14 | 李营 | Text semantic information analyzing system and method |
| WO2015062377A1 (en) * | 2013-11-04 | 2015-05-07 | 北京奇虎科技有限公司 | Device and method for detecting similar text, and application |
| CN105653649A (en) * | 2015-12-28 | 2016-06-08 | 福建亿榕信息技术有限公司 | Identification method and device of low-proportion information in mass texts |
| CN105653649B (en) * | 2015-12-28 | 2019-05-21 | 福建亿榕信息技术有限公司 | Low accounting information identifying method and device in mass text |
| CN107239447A (en) * | 2017-06-05 | 2017-10-10 | 厦门美柚信息科技有限公司 | Junk information recognition methods and device, system |
| CN107239447B (en) * | 2017-06-05 | 2020-12-18 | 厦门美柚股份有限公司 | Junk information identification method, device and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102567304B (en) | 2014-02-26 |
| JP5744228B2 (en) | 2015-07-08 |
| EP2657852A1 (en) | 2013-10-30 |
| EP2657852A4 (en) | 2014-08-20 |
| WO2012083892A1 (en) | 2012-06-28 |
| JP2014502754A (en) | 2014-02-03 |
| US20140013221A1 (en) | 2014-01-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102567304B (en) | Filtering method and device for network malicious information | |
| CN106598944B (en) | A kind of civil aviaton's security public sentiment sentiment analysis method | |
| Dey et al. | Opinion mining from noisy text data | |
| Hidayatullah et al. | Pre-processing tasks in Indonesian Twitter messages | |
| US10496749B2 (en) | Unified semantics-focused language processing and zero base knowledge building system | |
| Danisman et al. | Feeler: Emotion classification of text using vector space model | |
| US9898529B2 (en) | Augmenting semantic models based on morphological rules | |
| US20170286867A1 (en) | Methods to determine likelihood of social media account deletion | |
| CN105335352A (en) | Entity identification method based on Weibo emotion | |
| WO2019236360A1 (en) | Taxonomy enrichment using ensemble classifiers | |
| CN101599071A (en) | The extraction method of conversation text topic | |
| Sharma et al. | NIRMAL: Automatic identification of software relevant tweets leveraging language model | |
| CN107463703A (en) | English social media account number classification method based on information gain | |
| CN103473380A (en) | Computer text sentiment classification method | |
| Setiawan et al. | Feature expansion for sentiment analysis in twitter | |
| CN102779119B (en) | A kind of method of extracting keywords and device | |
| Dedhia et al. | Ensemble model for Twitter sentiment analysis | |
| Samonte et al. | Sentence-level sarcasm detection in English and Filipino tweets | |
| Tayal et al. | Fast retrieval approach of sentimental analysis with implementation of bloom filter on Hadoop | |
| Campbell et al. | Content+ context networks for user classification in twitter | |
| Banados et al. | Optimizing support vector machine in classifying sentiments on product brands from Twitter | |
| Sarna et al. | A probalistic approach to automatically extract new words from social media | |
| CN104281695A (en) | Combination theory based quasi natural language semantic information extraction method and system | |
| Morchid et al. | An author-topic based approach to cluster tweets and mine their location | |
| KR101240330B1 (en) | System and method for mutidimensional document classification |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20220620 Address after: 3007, Hengqin international financial center building, No. 58, Huajin street, Hengqin new area, Zhuhai, Guangdong 519031 Patentee after: New founder holdings development Co.,Ltd. Patentee after: Peking University Patentee after: BEIJING FOUNDER ELECTRONICS Co.,Ltd. Address before: 100871, Beijing, Haidian District Cheng Fu Road 298, founder building, 5 floor Patentee before: PEKING UNIVERSITY FOUNDER GROUP Co.,Ltd. Patentee before: Peking University Patentee before: BEIJING FOUNDER ELECTRONICS Co.,Ltd. |
|
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20140226 |